Kanna Idol - [Anima DiT]

Overview

A character LoRA trained natively on Anima Base v1.0 (Community DiT) capturing the facial structure, eye geometry, and styling of Kanna across multiple framing angles. Built with high concept disentanglement to keep facial identity locked while allowing full flexibility over outfits, poses, lighting, and backgrounds.

Usage Settings

• Trigger Word: kanna character
• Base Architecture: Community DiT (Kirazuri, Q-ANIMA, Z-Image, Anima Base v1.0)
• Optimal LoRA Weight: 0.80 – 0.90
  • 0.75 – 0.85: Flexible styling, highly receptive to radical costume and background changes.
  • 0.85 – 0.95: Maximum facial identity locking.
• Sampler: Euler_a
• Sampling Steps: 28 – 34
• CFG / Guidance Scale: 4.0 – 5.0
• Clip Skip: 1 or 2
• Target Aspect Ratios: 832 × 1216 (Portrait) or 1024 × 1024 (Square)

---

Recommended Prompting Templates

1. Standalone Character Generation (Clean Anime Aesthetic)

• Prompt: masterpiece, best quality, 1girl, solo, upper body, detailed expressive eyes, distinct facial features, soft natural lighting, depth of field, looking at viewer

• Negative Prompt: worst quality, low quality, score_1, score_2, score_3, bad anatomy, bad hands, missing fingers, extra digits, deformed limbs, disfigured, blurry, duplicate, cropped, watermark, text, signature, horns

---

Dual-LoRA Pairing Guide: Kanna + Cinematic Film Grain (2.5D Setup)

When stacking Kanna with the Cinematic Film Grain LoRA on Community DiT models (such as Kirazuri or Q-ANIMA), use the following weight profiles to balance anatomical identity with analog surface texture.

Recommended Weight Presets

• The 2.5D Sweet Spot (Recommended)

  • Kanna LoRA: 0.85
  • Film Grain LoRA: 0.50 – 0.55
  • Visual Outcome: Character facial contours, eye shape, and signature hairstyle remain firmly locked, while flat digital shading is replaced with micro-contrast, tactile clothing texture, and organic 35mm luminance grain.

• Clean Anime with Analog Softening

  • Kanna LoRA: 0.85 – 0.90
  • Film Grain LoRA: 0.35 – 0.40
  • Visual Outcome: Sharp linework and vibrant base colors are preserved, adding subtle film diffusion and muted specular highlights without altering line geometry.

• Gritty Cinematic / Vintage Emulsion

  • Kanna LoRA: 0.80
  • Film Grain LoRA: 0.65 – 0.70
  • Visual Outcome: Heavy photographic texture, textured shadows, and a faded analog aesthetic. (Avoid pushing Film Grain above 0.70 when Kanna is above 0.85 to prevent latent burn and high-contrast edge fringing).

Sampling Adjustments for Dual-LoRA Ingestion

• Trigger Stack: Place both activation tags consecutively at the beginning of the prompt: kanna character, film grain style, ...
• CFG / Guidance Scale: Set between 4.0 and 4.5 (Dual transformer adapters increase gradient pull; dropping CFG by 0.5 prevents highlight clipping and over-sharpened grain).
• Sampling Steps: 30 – 34 steps using Euler_a to allow the diffusion scheduler sufficient iterations to resolve grain in the dark tonal regions.